惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

量子位
WordPress大学
WordPress大学
小众软件
小众软件
云风的 BLOG
云风的 BLOG
IT之家
IT之家
人人都是产品经理
人人都是产品经理
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
Last Week in AI
Last Week in AI
博客园 - 【当耐特】
T
Tailwind CSS Blog
阮一峰的网络日志
阮一峰的网络日志
V
V2EX
宝玉的分享
宝玉的分享
博客园 - Franky
F
Fortinet All Blogs
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
GbyAI
GbyAI
Hugging Face - Blog
Hugging Face - Blog
Jina AI
Jina AI
D
Docker
博客园 - 聂微东
C
Check Point Blog
H
Help Net Security

Hacker News

GitHub - SeanFDZ/macmind: Single-layer transformer in HyperTalk for the classic Macintosh Show HN: Agent-cache – Multi-tier LLM/tool/session caching for Valkey and Redis Bonsai 1-bit WebGPU - a Hugging Face Space by webml-community Moving a large-scale metrics pipeline from StatsD to OpenTelemetry / Prometheus GitHub - Nightmare-Eclipse/RedSun: The Red Sun vulnerability repository GitHub - SethPyle376/hiraeth: Local AWS emulator focused on fast integration testing, with SQS support, SQLite-backed state, and a debug-friendly web UI. GitHub - macOS26/Agent: Any AI, replaces Claude Code, Cursor, OpenClaw. Over 18 LLM providers (Claude, OpenAI, Gemini, Ollama, Zai, HF, Qwen) wired into a native Mac app that writes code, builds Xcode projects, bumps versions, manages git, automates Safari, use AppleScript, JS or Accessibility, extend Agent! w/ MCP Servers, run tasks from your iPhone via Messages. YouTube now lets you turn off Shorts I Made a Terminal Pager Burgers | マクドナルド公式 Commands — HackerNews CLI documentation ChatGPT for Excel PiCore - Raspberry Pi Port of Tiny Core Linux Live Nation illegally monopolized ticketing market, jury finds Google Broke Its Promise to Me. Now ICE Has My Data. Founding Engineer at Adaptional | Y Combinator CRISPR takes important step toward silencing Down syndrome’s extra chromosome GitHub - saffron-health/libretto: The AI toolkit for building reliable browser automations US v. Heppner (S.D.N.Y. 2026) no attorney-client privilege for AI chats [pdf] Retrofitting JIT Compilers into C Interpreters IPv6 – Google The Accursèd Alphabetical Clock Cybersecurity Looks Like Proof of Work Now Fragments: April 14 Cal.com Goes Closed Source: Why AI Security Is Forcing Our Decision | Cal.com - Scheduling Software for Online Bookings Laravel raised money and now injects ads directly into your agent When moving fast, talking is the first thing to break Too much Discussion of the XOR swap trick – Heather Cafe Introduction to Spherical Harmonics for Graphics Programmers The Grand Line
GitHub - aattaran/deepclaude: Use Claude Code's autonomou...
2026-05-04 · via Hacker News

Use Claude Code's autonomous agent loop with DeepSeek V4 Pro, OpenRouter, or any Anthropic-compatible backend. Same UX, 17x cheaper.

Remote control running DeepSeek V4 Pro in the browser

What this does

Claude Code is the best autonomous coding agent — but it costs $200/month with usage caps. DeepSeek V4 Pro scores 96.4% on LiveCodeBench and costs $0.87/M output tokens.

deepclaude swaps the brain while keeping the body:

Your terminal
  └── Claude Code CLI (tool loop, file editing, bash, git — unchanged)
        └── API calls → DeepSeek V4 Pro ($0.87/M) instead of Anthropic ($15/M)

Everything works: file reading, editing, bash execution, subagent spawning, autonomous multi-step coding loops. The only difference is which model thinks.

Quick start (2 minutes)

1. Get a DeepSeek API key

Sign up at platform.deepseek.com, add $5 credit, copy your API key.

2. Set environment variables

Windows (PowerShell):

setx DEEPSEEK_API_KEY "sk-your-key-here"

macOS/Linux:

echo 'export DEEPSEEK_API_KEY="sk-your-key-here"' >> ~/.bashrc
source ~/.bashrc

3. Install

Windows:

# Copy the script to a directory in your PATH
Copy-Item deepclaude.ps1 "$env:USERPROFILE\.local\bin\deepclaude.ps1"

# Or add the repo directory to PATH
setx PATH "$env:PATH;C:\path\to\deepclaude"

macOS/Linux:

chmod +x deepclaude.sh
sudo ln -s "$(pwd)/deepclaude.sh" /usr/local/bin/deepclaude

4. Use it

deepclaude                  # Launch Claude Code with DeepSeek V4 Pro
deepclaude --status         # Show available backends and keys
deepclaude --backend or     # Use OpenRouter (cheapest, $0.44/M input)
deepclaude --backend fw     # Use Fireworks AI (fastest, US servers)
deepclaude --backend anthropic  # Normal Claude Code (when you need Opus)
deepclaude --cost           # Show pricing comparison
deepclaude --benchmark      # Latency test across all providers

How it works

Claude Code reads these environment variables to determine where to send API calls:

Variable What it does
ANTHROPIC_BASE_URL API endpoint (default: api.anthropic.com)
ANTHROPIC_AUTH_TOKEN API key for the backend
ANTHROPIC_DEFAULT_OPUS_MODEL Model name for Opus-tier tasks
ANTHROPIC_DEFAULT_SONNET_MODEL Model name for Sonnet-tier tasks
ANTHROPIC_DEFAULT_HAIKU_MODEL Model name for Haiku-tier (subagents)
CLAUDE_CODE_SUBAGENT_MODEL Model for spawned subagents

deepclaude sets these per-session (not permanently), launches Claude Code, then restores your original settings on exit.

Supported backends

Backend Flag Input/M Output/M Servers Notes
DeepSeek (default) --backend ds $0.44 $0.87 China Auto context caching (120x cheaper on repeat turns)
OpenRouter --backend or $0.44 $0.87 US Cheapest, lowest latency from US/EU
Fireworks AI --backend fw $1.74 $3.48 US Fastest inference
Anthropic --backend anthropic $3.00 $15.00 US Original Claude Opus (for hard problems)

Setup per backend

DeepSeek (default — just needs DEEPSEEK_API_KEY):

setx DEEPSEEK_API_KEY "sk-..."           # Windows
export DEEPSEEK_API_KEY="sk-..."         # macOS/Linux

OpenRouter (optional):

setx OPENROUTER_API_KEY "sk-or-..."      # Windows
export OPENROUTER_API_KEY="sk-or-..."    # macOS/Linux

Fireworks AI (optional):

setx FIREWORKS_API_KEY "fw_..."          # Windows
export FIREWORKS_API_KEY="fw_..."        # macOS/Linux

Cost comparison

Usage level Anthropic Max deepclaude (DeepSeek) Savings
Light (10 days/mo) $200/mo (capped) ~$20/mo 90%
Heavy (25 days/mo) $200/mo (capped) ~$50/mo 75%
With auto loops $200/mo (capped) ~$80/mo 60%

DeepSeek's automatic context caching makes agent loops extremely cheap — after the first request, the system prompt and file context are cached at $0.004/M (vs $0.44/M uncached).

What works and what doesn't

Works

  • File reading, writing, editing (Read/Write/Edit tools)
  • Bash/PowerShell execution
  • Glob and Grep search
  • Multi-step autonomous tool loops
  • Subagent spawning
  • Git operations
  • Project initialization (/init)
  • Thinking mode (enabled by default)

Doesn't work or degraded

Feature Reason
Image/vision input DeepSeek's Anthropic endpoint doesn't support images
Parallel tool use Disabled — tools execute one at a time
MCP server tools Not supported through compatibility layer
Prompt caching savings DeepSeek has its own caching (automatic), but Anthropic's cache_control is ignored

Intelligence difference

  • Routine tasks (80% of work): DeepSeek V4 Pro is comparable to Claude Opus
  • Complex reasoning (20%): Claude Opus is stronger — switch with --backend anthropic

VS Code / Cursor integration

Add terminal profiles so you can launch deepclaude from the IDE:

Settings > JSON:

{
  "terminal.integrated.profiles.windows": {
    "DeepSeek Agent": {
      "path": "powershell.exe",
      "args": ["-ExecutionPolicy", "Bypass", "-NoExit", "-File", "C:\\path\\to\\deepclaude.ps1"]
    }
  }
}

Or on macOS/Linux:

{
  "terminal.integrated.profiles.linux": {
    "DeepSeek Agent": {
      "path": "/usr/local/bin/deepclaude"
    }
  }
}

Remote control (--remote)

Open a Claude Code session in any browser — with DeepSeek as the brain:

deepclaude --remote                # Remote control + DeepSeek
deepclaude --remote -b or          # Remote control + OpenRouter
deepclaude --remote -b anthropic   # Remote control + Anthropic (normal)

This prints a https://claude.ai/code/session_... URL you can open on your phone, tablet, or any browser.

How it works

Remote control needs Anthropic's bridge for the WebSocket connection, but model calls can go elsewhere. deepclaude starts a local proxy that splits the traffic:

claude remote-control
  ├── Bridge WebSocket → wss://bridge.claudeusercontent.com (Anthropic, hardcoded)
  └── Model API calls  → http://localhost:3200 (proxy)
                            ├── /v1/messages → DeepSeek ($0.87/M)
                            └── everything else → Anthropic (passthrough)

Prerequisites

  • Must be logged into Claude Code: claude auth login
  • Must have a claude.ai subscription (the bridge is Anthropic infrastructure)
  • Node.js 18+ (for the proxy)

The proxy starts automatically and stops when the session ends. See proxy/README.md for technical details.

License

MIT